►
From YouTube: Kubernetes SIG-Window 20210914
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
All
right,
hello,
everybody
and
welcome
to
the
september
14th
2021
iteration
of
the
kubernetes
windows
community
meeting.
As
always,
these
meetings
are
recorded
and
uploaded
to
youtube
so
be
sure
to
adhere
to
the
cncf
code
of
conduct.
A
Let's
jump
in,
I
don't
think
we
really
have
any
new
announcements,
but
I'll
add
that
here
I
wanted
to
use
the
first
part
of
this
meeting
to
discuss
the
kep
status
for
123.,
so
we
actually
got
all
three
of
the
cups
that
we
were
tracking
in
for
123,
which
is
awesome,
great
work,
everybody
for
for
looking
at
that,
the
for
the
host
process
containers
kept
the
only
caveat
with
that
was
we
couldn't
convince
jordan
liggett
that
it
was
a
good
idea
to
update
the
some
of
the
kubernetes
client
libraries
to
be
able
to
understand
that
the
pathing
prefix
that
we
need
for
host
process
containers.
A
So
that
means
things
like
the
client
code.
Library
searching
for
the
like
service
account,
tokens
and
service
account
credentials
is
not
going
to
be
as
straightforward
as
possible.
We
do
have
a
lot
of
demos
of
how
to
work
around
that
issue
that
we'll
share,
but
other
than
that,
I
think
it
was
pretty
straightforward.
A
There
was
a
lot
of
either
suggestions
or
requirements
to
add
unit
tests
or
ed
tests
to
test
various
functionality,
which
I
think
will
help
improve
the
stability
and
usefulness
or
the
ease
of
use
of
these
and
kind
of
help
us
understand
what
scenarios
will
will
and
will
not
be
supported.
So
I
think
we're
kind
of
on
track
to
go
for
a
to
go
to
beta
in
123,
assuming
that
all
of
the
or
that
container
d
has
a
release
in
time
to
kind
of
support.
That
next
is
the
node
logs
this
one.
A
A
arvind
sebastian.
Do
any
of
you
guys
need
any
help
with
any
of
this.
A
Like
if
you
need
any
more
dev
resources,
now
might
be
a
good
time
to,
or
this
might
be
a
good
place
to
ask
for
help.
B
Yes,
definitely
helpful
be
welcome.
This
is
what
I
also
posted
on
the
community
slack
when
I
think
someone
had
asked
me
for
an
update
it
sort
of
stalled.
Both
me
and
christian
have
not
had
cycles
to
work
on
it.
Yet
so,
yes,
if,
if
if
there
are
other
folks
interested
in
this,
I'm
I'm
definitely
open
to
get
getting
help
here.
A
One
other
thing
to
note:
a
little
bit
different:
this
release
is
sig.
Node
is
imposing
a
soft
code
freeze
much
earlier
than
the
normal
code
freeze.
I
think
that
they're
wanting
any
implementation
prs
for
enhancements
to
be
open
to
have
the
pr
open
and
ready
for
review
by
october
15th
this
year
or
this
this
release.
So
that
is
something
to
note.
Yes,
I
did
see
your
the
thread
on
slack
as
well,
so
we
can
maybe
try
and
facilitate
extra
help
there.
On
slack.
C
A
A
D
A
Yeah
and
same
thing
for
this
cap,
any
of
the
node
changes
they'd
like
to
see
those
early
just
so
that
they
can
have
a
chance
to
fully
understand
how
all
the
different
like
enhancements
that
are
going
in
are
kind
of
interact
with
each
other
to
make
sure
there's
no
regressions
there.
But,
yes
thank
you
for
pushing
this
this
cup
to
alpha.
That's
a
lot
of
there's
a
lot
of
work
that
went
in
here
and
I
think
it'll
be
really
beneficial
down
the
line.
D
Yeah
no
problem
thanks
a
lot
everyone
in
the
community
like
sebastian,
you
jay,
like
all
of
us,
jumped
onto
a
call
last
week
and
to
get
the
approval.
So
thanks
a
lot.
Everyone.
A
E
Yeah
we
had
some
when
we
had
some
questions
so
on
the
host
process,
containers
that
she
was
hitting
she's
in
the
beijing
time
zone.
So
I'm
kind
of
just
being
the
merchant
here
and
I
figured
maybe
we
could
look
at
this
together.
If
anybody
had
any
idea,
it
looks
like
you
already
looked
at
it
more,
but.
E
A
Run
this
without
seeing
the
cluster
that
might
be
oops,
okay,
difficult
but
yeah.
I
just
see
that
when
they
did
pass
the
test
args
through
and
that's
what
I
was
suggesting-
that's
what
we
had
to
do.
E
Yeah,
okay,
well,
okay,.
A
Yeah
this
so
one
thing
no
th
like
okay,
all
the
test
args
should
just
get
passed
to
the
ede
test,
binary
itself,
but
sometimes
there's
not
a
one
direct
mapping
of
what
you
specify
in
the
the
pro
job
to
the
ede
test,
args,
because
this
kind
of
gets
translated
by
the
container
that's
running
and
do
that
you
can
take
a
look.
Is
there
a?
I
didn't
see
a
slack
thread
about
this.
That
might
be
a
good
place
to
start,
but
yeah
I'll
take
a
deeper
look
after
the
meeting.
A
A
Yeah,
so
we
do
have
that
that
feature
skipper.
Oh
maybe
it's
getting
skipped
between
node.os
distro
is
not
set
to
windows.
E
A
A
A
Yeah
that
that
should
be
enough
and
make
sure
that
you're
running
container
d
from
the
master
branch
not
a
released
version,
because
you
need
to
have
the
hcs
that
either
you
need
to
either
have
the
hcsm
version
that
understands
the
annotations
that
we
pass
through
for
the
alpha
implementation
on
the
cri
calls
or
container
d
that
understands
the
nuke
cri
fields
and
both
of
those
are
only
kind
of
available
if
you
build
from
container
d
or
h
system
from
master.
But
this
is
a
good
package
to
build.
A
We
built
this
out
of
our
or
which
one.
A
E
A
A
Is
the
the
container
d
build
that
we
have
so
we've
updated
the
one
of
the
kubernetes
projects
to
have
a?
Let
me
that
might
be
worth
highlighting
here.
E
A
So
in
sig
windows
tools:
now
we
have
a
github
action
that
we
have
a
workflow
that
will
build
container
d
nightly
and
publish
it
to
so.
We
build
hcs
gym
from
master.
We
build
and
continue
d
and
ctr.exe
from
master
of
container
d,
and
this
runs
nightly
and
ends
up
as
a
release,
and
so
this
release
here
will
get
updated
every
every
night
with
the
latest
build.
So
this
is
a
good
if
you're
interested
in
testing
host
process
containers
or
just
ensuring
that
there's
no
regressions
in
continuous
d
for
windows.
A
This
is
a
good
container
d
package
to
target.
Oh.
A
A
So
this
we
don't
need
to
go
into
the
details
of
this,
but
that
will
kind
of
ease
testing.
A
I
added
this
one
jay.
This
might
be
related
to
what
we're
just
talking
about
earlier,
but
I
saw
this
issue
just
got
closed
due
to
inactivity.
It
looks
like
there
was
a
pr
to
address
this,
but
oh
yeah.
It
wasn't
clear
that
this
is
actually
addressed.
E
A
A
Okay,
does
anybody
else
have
anything
they
want
to
discuss
before?
I
guess
we
jump
into
windows,
node
conformance,
I
think
that'll
probably
take
up
the
rest
of
the
agenda
today.
A
A
E
A
meme
reissued,
it
I
closed
mine,
and
then
we
did
it.
It
means
account
there
we
go
here.
It
is
so
this,
as
is,
is
pretty
opinionated,
but
it's,
I
think,
like
minimal.
I
think
this
is
like.
F
E
Okay,
would
you
like
to
be
a
co-author
on
this
with
us
by
the
way
sure,
okay
great,
so
I'm
going
to
commit
these
and
then
I'm
just
going
to
like
give
your
give
us
your
get?
Well,
I
have
it
right
here
right
so
meme
are
you
here.
J
E
K
Yeah,
I
said
the
first
thing
I
said
was:
which
versions
do
we
want
to?
Support
was
my
first
kind
of
comment,
and
what
do
we
consider?
K
A
E
E
A
A
F
F
Oh
sorry,
go
ahead.
Go
ahead,
no
go
ahead.
2004,
I
believe,
is
about
two
months
away
from
being
end
of
service.
So
I
I
don't
know
the
time
frame
for
getting
this
in,
but
if
we
say
2004
and
then
by
the
time
we're
like
have
all
the
tests
written
2004's
end
of
service.
The
issue
we've
had
on
the
rancher
side
is
those
amis.
They
just
disappear,
so
we
don't
want
to
go
through,
adding
all
the
tests
in
for
all
these
things
and
then
there's
no
images
to
actually
use
to
test
for
them.
F
K
Yeah
I
feel
like
there
was
a
bigger
issue
going
from
like
2019
upwards,
because
there
was
a
lot
of
stuff
kind
of
trickling
in
for
sort
of
2004
in
terms
of
container
support,
but
I
think
now
you're
kind
of
on
the
2022
lcsc.
I
think
most
of
those
issues
have
kind
of
gone
away
for
now,
and
I
guess-
and
so
now
it's
almost
like-
we
could
almost
just
say
we
are
only
going
to
conformance
against
the
the
ltscs
like
the
last
two
maybe
and
forget.
K
C
Think
it's
fair.
I
think
it's
fair
to
do
20
h2,
but
to
to
ross's
point.
Yes,
the
2004
is
going
to
be
out
of
service
december.
14Th
is
basically
exactly
three
months
from
now
and
I
I
think
it
doesn't
make
much
sense
to
do
2004
at
this
point.
So
I
think
two
ltsc
and
one
sec
may
be
the
right
call,
because
20
h2
is
gonna
be
available
until
august
2022.
A
One
kind
of
question
I
had
about
this
is:
do
we
think
that
there's
a
need
to
call
out
which
windows
server
versions
are
actually
supported
in
in
something
like
this?
Like
people,
don't
say
that
you
know
different
either
ltsc
or
the
faster
releases
of
like
ubuntu
or
anything
are
conformant
or
not
things
just
work
and.
A
Is
different
and
I
do
know
that
we
need
to
have
the
pause
images
although
or
like,
although
that
is
changing
as
well
with
the
enhanced.
I
forgot
exactly
what
it's
called,
but
the
enhanced
compatibility
for
process,
isolated
containers
that
brendan
discussed
one
of
the
previous
meetings,
so
that
that
will
hopefully
help
improve
the
story
here
too.
But
I'm
wondering
if
there's
actually
a
need
to
certainly
like
say,
different
os
versions.
F
E
Actually
so
this
is
not
this,
this
actually
you're
right.
So
this
doc
is
not
about
what
we
what
we
test
it's
about,
what
conforms,
and
so
it
actually
could
be
os
neutral
right.
Just
the
same
way,
the
actual
the
actual
yeah
we
don't
sit
around
and
talk
about.
1809
ubuntu
has
node
port
connectivity
like
I've,
never
heard
anybody
say
that
before.
A
E
I
think
the
way
this
is
written
and
I'm
not
at
all
married
to
the
way
this
is
written,
but
in
the
way
that
this
is
written,
the
idea
was,
we
probably
need
to
change
deeds
from
being
an
approver.
It
says
that's
an
api
machinery
thing
I
guess,
but.
A
Kind
of
all
over
the
place-
and
I
believe
that
david
has
been
the
pr
reviewer
for
most
of
the
recent
windows
caps
as
well.
So,
okay,
he
I
think
it
I
don't
know
if
it
would
be
better
or
worse,
to
keep
him
as
the
pr
reviewer
on
one
hand,
he's
getting
more
and
more
familiar
with
the
nuances
of
windows
versus
linux
containers.
On
the
other
hand,
it
might
be
better
to
you
know,
spread
that
across
more
of
the
kind
of
architectural
type
people.
I.
E
Feel
like
being
serialized
on
someone
from
api
machinery
is
not
necessary
for
this,
because
he's
a
he's,
a
busy
man,
I
mean
that's.
A
H
Well,
anyways:
we
can.
We
can
think
about
that,
but,
like
yeah
yeah,
I.
D
Okay
but
jay,
I
think
one,
oh.
E
Busy
but
yep.
D
E
A
Well,
the
os
the
host
process
stuff
should
work
on
like
it'll
work
on
2019.
That's
where
all
of
our
tests
are.
So
it's.
E
Yeah,
well,
not
only
will
it
work
on
2019,
but
also
it's
a
separate
thing.
So
I
separated
the
host-
oh
okay,
from
core
right,
so
it's
like
here's,
the
core
stuff
yeah,
and
then
this
should
work
anywhere
and
then
the
host
process
stuff.
It's
like
that's
a
separate,
and
I
don't.
I
don't
know
how
I
feel
about
the
way
we've
separated
these,
but
you.
E
A
theoretical
starting
point
I'm
like
well,
you
could
envision
a
world
where
you
had
a
sona
belief,
flag
where
you
said
windows
conformance
and
then
you
could
give
it
sub
options
of
the
different
things
that
you
cared
about
right
and
I'd
put
active
directory
in
host
process
of
those.
But
as
just
as
examples,
but
for.
E
C
I
think
we
should
think
this
through.
Mark
I
mean
dual
stack
is
one
example.
I
think
we
have
a
couple
more
example
in
networking
where
it
is
os
specific.
So
if
the
the
purpose
of
the
conformance
test
is
to
give
confidence
that,
if
somebody
is
running
a
distro-
and
you
know
they
feel
that
you
know
all
these
test
cases,
whatever
will
be,
you
know,
or
you
know,
are
running
well,
I
think
in
windows.
Unlike
linux,
it
has
dependency
on
os,
because
the
os
version
will
have
some
capabilities
that
the
other
won't,
and
this
will.
C
Beyond
2022
right
so,
let's
say
even
dual
stack
right:
let's
say
you
have
a
test
or
or
something
for
conformance
there
and
you
test
it
for
2022
and
then
you're
like
well.
My
destroy
is
conformant
for
2019,
20,
h2
and
every
you
know
2022.
That's
not
really
the
case
right.
It
is
conforming
for
2022.
It
is
not
conforming
for
other
os
versions.
So
I
think
if
the
purpose
of
conformance
is
to
give
confidence,
I
I
think
we
have
to
be
specific
in
windows
cases
here.
A
E
M
E
E
All
right,
here's
the
deal
so
like
think
of
it
this
way,
like
the
purpose
of
conformance.
So
I
you
could
challenge
the
premise
of
that
right,
like
you
could
challenge
the
idea
that
the
purpose
of
conformance
is
to
is
to
verify
that
an
operating
system
is
correct.
The
purpose
of
conformance
is
to
act
to
verify
that
a
cluster
is
conformant
to
the
kubernetes
api
specification
right.
So
at
that
level,
decoupling
from
the
os
might
be
might
be
valuable.
B
B
B
I
don't
think
our
case
really
applies
here.
It's
a
little
special
because
we're
using
this
hybrid
overlay
piece
inside
windows.
That
requires
a
feature
that
was
added
to
the
kernel,
but
it
from
what
I
understand
there's
no
way
it's
going
to
make
its
way
into
2019,
because
2019
is
very
you
know
the
the
rate
of
stuff
being
backboarded
is
very
slow
as
what
I
what
I
was
told
so
could
you
verify.
A
I
was
going
to
say:
does:
does
the
patch
you
need?
Is
it
needed
for
functionality
that
openshift
is
specifically
is
relying
on
or
is
it
needed
to,
or
is
it
needed
for
like
saying
that
the
cluster
is
conformant?
B
B
In
windows
right,
you
could
end
up
with
an
sac
that
has
a
feature
that
is
needed
for
kubernetes
to
work,
but
not
anymore,
yeah
not
anymore.
Hopefully,.
C
But
but
but
there
will
always
be
cases
mark,
I'm
saying
I
mean
there
will
be
cases
and
I
can
call
out
specific
cases
and
jay
to
your
point
right.
C
I
mean
yes
you're
looking
at
cluster
and
that
is
it
conformant
to
the
kubernetes
api
or
not,
but
in
some
cases
it
won't
be
because
of
the
os
limitation
right
and-
and
if
that's
the
case,
then
the
plus
you
know
you're,
looking
at
just
the
cluster
level
from
an
api
perspective,
but
itself
the
cluster
won't
be
able
to
support
it
because
the
os
doesn't
have
that
functionality.
E
A
A
Reasoning
is
the
service
or
yeah.
The
servicing
team
has
these
requirements
where
they
they
won't
release
a
fix
to
a
version
of
windows.
If
it's
not
already
present
in
all
versions
of
windows
released
after
that,
and
then
there
was
always
some
requirements
for
they.
They
call
it
flighting,
but
to
have
the
fixes
of
like
available
in
to
to
customers
to
a
subset
of
customers
to
use.
So
if,
if
not
because
we
had
so
many
different
sac
releases
in
the
windows,
server
2019
code
base
was
so
old.
A
If
there
was
an
issue
that
was
fixed,
it
was
required
to
be
fixed
in
the
most
recent
sac
release
and
then
the
most
recent
then
the
previously
released
tech
release
and
so
on
and
so
on.
Until
it
hit
windows,
server,
2019,
but
now
windows,
server
2022
is
going
to
be
the
minimum
like
the
the
latest
version,
so
that
will
be
the
first
windows
server
version
where
the
fix
will
be
available
so
that
we're
super
excited
about
that
too.
C
A
D
I
think
one
thing
I
wanted
to
add
is:
where
should
we
call
out
that
the
conformance
test
is,
is
something
that,
like
a
particular
test
suit,
it's
going
to
work
on
a
particular
version
of
windows?
D
The
way
I'm
seeing
about
it-
and
I
could
be
wrong-
is
this
should
be
somewhere
within
the
test
like,
for
example,
if
I'm
running
the
gmc
gms
test
or
some
other
test,
I
would
like
a
particular
version
of
windows
like
list
all
the
nodes
that
are
available,
find
out
the
build
version.
That
is
that
this
particular
test
can
run
on
and
if
the
not
the
notes
are
not
having
that
particular
build
version,
then
we
need
to
stop
running
the
test.
We
need
to
skip
it
at
that
point
of
time.
A
E
That
I
mean,
I
think,
that
they're
they're
that
goes
back
to
the
whole
like
what
should
the
tests
do,
and
I
think
that
the
bar
here
that
I
kind
of
started
with
was
was
that
the
tests
should
have
a
it's
this.
It's
a
boolean
output
from
the
tests
right,
as
opposed
to,
like
you
know,
a
variable
thing,
and
then
you
can
customize
the
granularity
of
that
boolean
output
and
what
it
means
by
by
adding
other
subsuites
of
verification
of
things.
E
And
then
I
guess
you
could
presumably
see
a
tool
like
sonaboo
or
something
saying
if
a
user
is
running
all
of
their
nodes
on
1809
and
they
try
to
do
host
process
containers,
I'm
going
to
blow
up
or
whatever
you
know,
but
I
see
that
as
like
in
the
user
or
the
testers,
like
that's
part
of
their.
I
feel
like
a
big
mistake.
We've
made
in
kubernetes
is
the
entire
thing
of
how
we
skip
if
os
distro
is.
I
feel
like
that.
Didn't
work
out
very
well
like,
like.
E
D
N
N
Sorry
I
was
mentioning
that
that
that
script
logic
is
only
for
non-conformance
tests
by
the
way,
for
example,
for
common
conformance
tests
or
not
conference
tests.
N
We
also
have
the
lyrics
money
tag,
so
we
exclude
through
that
tag,
then,
is
that
what
kubernetes
decided
that
there
should
be
no
skip
logic
inside
the
tests
right?
So
there
is
the
compromise
that
there
are
times,
as
only
that
we
skip
them
in
our
test
runs.
You
know
yeah,
but
yeah.
N
E
I
mean
the
the
reason
that
it
exists.
This
way
today
is
that
we
never
really
figured
out
how
to
properly
vendor
a
testing
library,
because
we
were
kind
of
building
the
end-to-end
tests
as
we
were
using
them
right,
so
you're
kind
of
just
copying
and
pasting
code
from
different
things
and
so
like.
E
If
we
were
to
design
the
test
from
scratch
today,
I
don't
think
we
would
have
sort
of
had
a
monolithic
testing
package
and
then
had
bespoke
logic
in
it.
Right
like
we
would
have.
We
would
have
probably
thought
more
about
how
to
generically
define
the
what
we
cared
about.
You
know
in
a
way
that
was
like
container,
independent
and
and
implementation
independent
and
there's
a
lot
of
stuff
going.
You
know
it
means
like
so
I
mean
I
think,
we've
done
an
okay
job
with
that,
but
you
know.
E
Like
actually,
I
think,
robbie's
pr
is
kind
of
important
for
that
right,
like
the
ability
to
actually
define
the
os
in
the
in
the
container
itself
is
going
to
actually
improve
the
quality
of
our
tests.
A
lot,
I
think,
but
yeah,
that's
a
tricky
one.
I
I
don't,
I
don't
know
ravi.
I
don't
know
what
the
answer
to
that
is,
but
you
know
if
you
want
again
like
I
think
if
you
want
to
co-author
this
with
us,
you're
welcome
to
right.
E
If
you,
if
you
want
to
add
that
story
or
if
you
just
want
to
add
a
comment
and
like
review
it,
that's
that's
totally
cool
too.
I
I
think
it's
tricky
to
think
about
that.
For
me,
though,.
D
I
don't
I
don't
know
so
to
be
clear
to
you.
What's
the
time
frame
that
we
are
looking
at
for
this,
because
I
think
this
release,
I
will
be
busy
working
on
the
os
in
the
thoughts
back
first.
E
Well,
well,
you
know
me
robin
I'm,
not
very
patient,
so
whatever
I
mean,
let's
get
it
done.
I
mean
if
there's
something
important,
that
we're
missing:
let's,
let's,
let's
fix
it,
but
other
than
that.
I
I'm
not
I'm
not
going
to
sit
here
and
like
bang
down
doors,
to
get
it
into
123,
but
I
feel
like
we
should.
Probably
you
know,
get
everything
we
care
about
into
this
thing
and
then
work
out
the
details
as
a
sig
quickly
and
then
you
know.
D
Yeah
at
high
level
to
me
conformance,
is
something
that
is
at
least
os
version
agnostic
in
case
of
windows.
Rest
of
the
tests
say
if
it
is
actually
depending
on
a
particular
version
of
windows
or
a
particular
version
of
container
d,
or
something
like
that,
then
perhaps
that
should
not
become
a
conformance
test,
is
what
I'm
thinking.
E
Yeah,
you
know
jace
demars
mentioned
this.
We
talked
to
him
yesterday
and
it
was
like
well.
He
was
talking
about
this
distinguished
from
the
idea
of
like
maybe
distinguishing
conformance,
because
also
this
that
could
just
start
a
war
if
we
say
conformance
right,
like
yeah,.
E
N
So
I
agree
with
ravi
when
he
mentioned
that
it
shouldn't
depend
on
the
os
version.
I
think
that's
a
fair
point,
but
in
regards
to
the
continuity
version
less,
so
it's
a
lot
easier
to
update
container
d
than
to
update
the
entire
operating
system.
In
my
opinion,
and
additionally,
this
another
thing
that
has
been
bugging
me
for
quite
a
while.
N
So
let's
take,
for
example,
gmsa
you
actually
have
to
have
particular
features
enabled
or
set
up
in
order
to
actually
use
it.
But
we
still
want
to
include
that
in
the
windows,
conformance
definition
right,
because,
typically,
if
you
are
to
somehow
run
windows
conformance
with
sono
boy
or
something
like
that
to
be
mentioned
in
the
spec
block,
someone
would
have
to
know
explicitly
that
they
have
to
enable
that
or
how
to
set
it
up.
So,
if
you're
talking.
N
N
Yeah
but,
for
example,
host
process
is
still
in
alpha
right
right,
but
host
process.
N
This
in
this
spec
here.
D
I
need
to
drop,
but
I
think
I'd
like
to
continue
the
discussion,
perhaps
on
different
stack.
E
Yeah,
okay,
ravi
just
ping
us
and
let
me
know
if
you
want
to
help
us
author
this
and
be
honest
with
us
and
also
youtube
clients.
E
Yeah,
claudia,
let's
talk
about
it
later,
just
ping
me
because
I
think
you're
offline,
obviously
ross,
please
just
push
a
pr
up
there
and
if
you
want
to
be
an
author
with
us,
you're
welcome,
of
course,
you're
also
a
reviewer.
If
you
prefer
to
do
that
so,
but
I
mean,
I
think,
it's
great
for
us
to
have
different
people
from
different
companies
doing
this,
like.
I
don't
want
a
person
that
you
know
pushes
this
so
yeah
a
memes.
E
This
is
a
memes
repo,
so
anybody
interested
in
contributing
just
just
send
them
in
your
github
and.
E
We'll
go
from
there
cool
perry.
Can
you
add
the
do
you
do
you
want
to
add
the
conclusion
to
that
or
do
you
think
we
even
reached
a
conclusion
on
the.
K
E
All
right,
you
want
to
push
that
up
or
yeah
somewhere,
okay,
cool,
so
I'll
tag
you
on
that
and
then
I
feel
like
we've
kind
of
bled
into
pairing.
So
I
feel
like
this.
Probably
we
can
kind
of
just
end
and
with
this.
E
I
think
that
kind
of
addresses
some
of
the
stuff
that
folks
have
been
saying
about.
Like
you
know,
privileged
containers
and
stuff,
like
I
think
the
thing
to
keep
in
mind
is
the
core
conformance
part,
I
think
that's
kind
of
like
we
expect
this
to
work
anywhere.
I
think
this
is
kind
of
to
me.
The
most
important
part.
E
J
Doesn't
work
so
jay
how
much
of
that's
already
implemented
of
this
test
suite
most
of
the
core
stuff.
E
Is
the
only
stuff
that's
not
implemented
is
so
I
suggested
these,
but
these
aren't
implemented
nobody.
I
thought
people
would
like
argue
with
me
about
these,
but
nobody's
complained
about
these
either
because
nobody's
read
them
yet
or
because,
but
like
I
feel,
like
we've
hit
things
me
and
perryfit
stuff,
where
container
d
is
weird
on
windows
and
logs
get
accumulated
and
stuff.
E
Sometimes
I
feel
like
the
service
proxy
on
windows
is
not
as
well
exercised,
and
so
I
would
like
to
have
like,
especially
when
you
think
of
large
clusters
right,
like
a
hundred
node
windows
cluster.
If
someone
ever
make
one,
I
I
think
it
would
be.
E
I
think
the
the
reason
I
think
it's
important
to
have
these
service
tests
in
place
is
that
we
can
say
like
look.
Okay,
if,
for
some
reason,
the
windows
coupe
proxy
doesn't
support
10
000
services,
then
if
somebody
was
to
try
to
run
a
thousand
node
cluster,
these
tests
would
fail,
and
that
would
give
us
a
good
way
to
say
you
can't
run
a
windows
cluster
of
greater
than
this
size,
because
I
don't
think
any
of
us
really
knows
how
all
this
hms
proxy
stuff
works
at
scale.
So
I.
I
F
Running,
like
I
mean
that's,
I
think
it
was
like
116
117
kubernetes,
but
it
just
was
not.
It
was
not
pretty
yeah,
it's
also
all
bare
metal,
of
course
giant.
You
know
mega
nodes
which
doesn't
help
anything
yeah
I
mean.
Have
we
ever
looked
at
like
scale
testing
in
the
sig
windows?
No,
no!
Well,.
F
I
think,
because
I
think
that's
the
next
step
after
conformance,
is
taking
the
conformance
building
test
and
scaling
up
with
the
conformance
test
until
we
hit
a
point
where
something
breaks
and
then
saying
you
know
documenting
that
and
then
just
leaving
it
there,
it's
not
something
we
should
have
to
repeat
too
often
either,
because
windows
is
slow
in
terms
of
development,
yeah,
expensive.
E
J
J
E
E
N
E
Yeah,
I
think
that
for
me
it's
the
I'm
not
too
worried
about
the
pod
rotation,
I'm
more
worried
about
the
service
load,
balancing
part,
because
I've
always
seen
that
that's
the
thing
that
breaks
right
like
it's,
that
host
local
load,
balancing
stuff
that
never
seems
to
work
properly
but
yeah.
A
So
the
I
missed
a
whole
bunch,
but
the
though
there
was
a
recent
effort
to
define
a
bunch
of
slos
and
slis
for
kubernetes
and
for
linux
they
define
the
number
of
pods
that
can
run
on
a
node
based
on
the
number
of
cores.
So
I
think
that
they
say
it's
yeah
minimum
or
it's
it's
the
max
of
some
number
or
like.
I
think
it's
10
times
the
number
of
cores.
It
might
be
helpful
to
define
that
in
that
regard,
as
well.
K
E
So
I
think
you
can
run
yeah,
so
it's
like
the.
I
didn't
put
anything
per
four,
because
I'm
assuming
that
even
a
minimal
cluster
should
be
able
to
run
10
containers,
and
so
then
the
question
winds
up
being.
Can
you
load
balance?
If
you
have
a
hundred
nodes,
can
you
load
balance
a
thousand
pods
behind
those
hundred
nodes.
A
I
wasn't
they
they
did
do.
I
think,
like
they
I'll,
try
and
link
to
those
two.
They
did
do
like
some.
They
did
kind
of
break
it
up
and
say
these
were
cluster-wide
like
goals,
and
these
were
node-wide
goals
might
be
good
to
do
something
like
that
like
on
on
windows,
some
more
I
mean,
as
always,
but
even
more
in
windows,
the
more
stuff
you
have
running
on
any
node
stresses
things
like
network
programming,
more
so
than
on
linux.
If
yeah.
A
Happen
on
the
node,
but
yes,
I
think
we
should
just
try
and
move
in
the
direction
that
sig
scalability
is
using
for
how
they
define
some
of
their
like
metrics
or
recommendations.
E
Well,
but
here's
the
thing,
though,
the
one
thing
here
is:
I
I
I
I
was
I
intentionally
steered
away
from
scalability
because
the
I
think
the
funk,
the
purpose.
What
I
was
thinking
the
purpose
was
here,
is
to
define
functionality
so
by
running
10
pods
behind
a
cluster
ip
you're,
not
really
saying
much
about
scale.
E
C
E
A
I
don't
know
I'm
not
saying
this
yeah,
I
could
see
all
of
that.
I
missed
some
of
it.
I
think
if
we
want
to
say
like
in
order
to
make
sure
things
are
working
like
what
you
said,
we
want
to
have
the
test
case
at
least
cover
this
minimum
scenario,
but
not
necessarily
talk
for
how
this
scales,
I'm
completely
fine
with
that
that
makes
sense.
E
Yeah
yeah,
it's
like
basic,
it's
like,
and
it's
like
my
fault
because
I
said
before
I
should
have
said,
like
basic
functionality,
concurrent
functionality
right,
it's
like
that's
the
idea
right.
A
N
E
K
E
J
E
K
And
then
one
thing,
the
one
thing
I
was
just
as
we
were
all
talking
I
was
thinking
about
was
we
started
talking
about
minimum
like
patch
level
and
stuff?
Do
we
as
a
specific
windows
want
to
come
up
with
a
list
of
patches
that
we
say
we're?
Not
you
know
it's
not
conformant,
or
at
least
operationally
conformant.
K
A
A
F
Around
yeah
and
we
can
always
wrap
some
like
you
know
in
today's
current.
Whatever
you
know,
cvs
popping
all
the
time,
especially
zero
days
like
I
think
three
months
is
generous
yeah
like
if
you're
missing
three
months,
that
that's
a
problem.
That's
not
our
problem.
E
Maybe
this
needs
to
be
moved
out
or
something,
and
I
wonder
whether
we
should
consider
that
being
like
a
thing
that
we
say
we're
going
to
curate
as
part
of
this
cap.
You
know
like
because
it's
going
to
change
over
time
right,
so
it's
not
part
of
the
specification.
So
much
as
it
is
like
an
ongoing
thing
that
makes
the
specification
valuable.
E
Is
that
now
your
way
of
saying
these
things
actually
work
or
not,
and
then
you'll
curate,
that
list
of
data
and
then
maybe
where
we
create
that
list
of
data
is
part
of
this
gap
as
well
right.
So
like
okay,
we'll
manage
a
whatever
a
readme
file
in
sig
windows,
blah
blah
blah
that
will
have
up-to-date
information
about
these
blah
blah
blah
things.
Yeah.
E
Yeah
we
did
that
for
the
policy
stuff
we
did
that
like
we
were
like
okay,
here's
how
the
tests
are
going
to
work.
They
do
this
right
like
and
then,
and
that
took
a
long
time
to
implement,
though
that
took
us
a
year
and
a
half
to
get
that
stuff,
working
and
proud
properly
merged
but
like,
but
once
it
was
in
it
was
nice.
So
I
think
it
would
be
good
to
specify
the
goals
the
runtime
goals,
but
I
think
we
should
be
honest
when
we
do
that,
because
we're
so
strapped
for
cycles.
E
K
Oh
yes,
claudia,
I
gotta
run
guys
I'll
catch
you
tomorrow,
jay
and
I'll
talk
to
you
all
next
week,
yeah.
E
Yeah
claudio
add
yourself
either
as
an
off
and
just
push
some
commits
to
this
and
just
give
us
your
github,
or
I
think
we
have
it's
claudio
blue
and
just
push
some
commits
up
for
this.
If
you
want,
and
if
you
want
add
yourself
as
a
reviewer,
if
you
don't
want
to
be
an
author
on
it,
just
if
you
want
to
add
yourself
as
a
reviewer,
then
you
know
that
would
be
cool
too
yeah
I'll.
Take
a
look
too,
I
think,
mark
you
already
said:
you're
going
to
be
a
reviewer,
though.
A
E
O
A
A
H
A
E
So
so
we
actually-
and
it's
because
that's
also
partially
from
talking
to
jace
to
mars
about
this
yesterday.
He
mentioned
the
same
thing
and
the
idea
being
that
like
well.
Maybe
if
we
can
get
this
working
at
the
windows
level,
we
can
start
to
broaden
the
def,
broaden
the
types
of
certifications
we
have
for
kubernetes,
so
that
we're
not
serialized
on
this
one
definition
of
conformance,
because
it's
kind
of
starting
to
that's
starting
to
break
that
whole
model.
Right
like.
A
Yeah
and
like
conformance
is
to
me,
is
meant
to
prevent
vendor,
locking
it's
like,
and
that's
the
only
thing
that
it's
meant
to
do
and
like
internally
here
everybody
says:
oh,
did
you
run
the
conformance
tests
and
I
always
have
to
cur
well
that
just
like
sets
me
off
and
I'm
like.
No,
you
need
to
ask
if
they
ran
the
appropriate
ede
tests,
because
performance
tests
don't
test
what
don't
necessarily
test
what
you
think
they
do.
C
E
A
E
Windows,
like
has
this
yeah,
like
so
with
a
windows
cluster
that
has
linux
nodes
on
it,
should
pass
conformance,
and
that
has
meaning
right
and-
and
I
think
that
that's
like
that's
what
the
conformance
definition
is.
It's
telling
you
that
that
those
300
or
so
tests
pass,
some
of
which
are
linux,
specific
right,
and
it's
always
going
to
be
what
that
means,
and
so
we're
not
skirting
it.
What
we're
just
saying
is
that
like,
in
addition
to
passing
conformance,
which
tells
you
a
lot,
you
there's
also
this
other
suite
of
tests.
E
E
H
A
J
E
To
be
a
plugable
sort
of
library
for
like
for
running
a
battery
of
like
informing
tests
on
your
cluster,
and
I
think
that
the
sort
of
poster
used
the
quintessential
usage
of
it
is
certified.
Conformance,
okay,
you
know
but
well
you
know,
you
know
we're
good
friends
with
vlad
over
here
and
you
know
we
could
just
we
can.
We
can
he's.
Definitely
I
mean
we're
all
in
the
business
of
of
making
sona
bui
capable
of
doing
more
than
that,
but
I
mean
he's,
certainly
not
opposed
to
that
right
I
mean
certain.
E
Certified
performance
is
just
that's
just
one
of
of
the
use
like
normally
when
I
run
sonic
buoy
I
use
e3
focus,
equals
sig
network
or
something
or
sig
windows.
You
know
so
so
yeah,
but
I
mean
it's
a
good
point
like
I
think
that
has
to
be
clarified
in
sono
boy
that,
like
this,
you
know-
and
I
think
actually
in
the
way
that
I
implemented
the
sono
buoy
example.
I
think
I
kind
of
broke
the
rule
that
I
just
said
here
because
see
I
said,
windows
performance.
I
really
should
have
said
something
else.
I
E
E
I'm
gonna
update
that
sauna
buoy
thing
as
well.
E
E
E
Yeah,
I'm
just
updating
all
these,
so
I'm
gonna
go
down
to
my
sonogui
here
and
I'm
gonna.
Finally,
here
I
am
so
it's
like,
so
we
plug
in
install
and
we
can
do
like
windows.
E
E
Okay,
so
that's
cleaned
up,
so
I've
removed
a
lot
of
the
conformance
stuff.
Well,
thanks
mark
for
rejoining
and
ross
and
everyone,
so
everyone
feel
free
to
mob
on
it
and
just
get
involved.
I
G
G
E
Right
anything
else:
anybody
wants
to
go
through
free
drink.
You
got
anything
no
just
hanging
in
there
all
right
cool,
I'm
gonna
jump,
I'm
gonna
get
something
to
eat
and
then
we'll
sync
up
we'll
be
around
later.